A. Understanding the Distribution of Sales

ggplot(sales_data, aes(x=Global_Sales)) + geom_histogram(fill='navy', color='lightblue') + ggtitle('Global Sales - Histogram (bins=50)')

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

ggplot(sales_data, aes(x=Global_Sales)) + geom_histogram(bins=100, fill='navy', color='lightblue') + ggtitle('Lowest Global Sales - Histogram (bins=100)') + xlim(0, 5)

## Warning: Removed 204 rows containing non-finite values (`stat_bin()`).

## Warning: Removed 2 rows containing missing values (`geom_bar()`).

# log transformation on Global Sales
ggplot(sales_data) + geom_histogram(aes(log(Global_Sales)), fill='navy', color='lightblue')

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Inference:
From the above histogram plots, we can see that the data is highly skewed and mostof the games have sales less than 1 million dollars, hence while fitting the models we should Log transform the column.

B. Understanding the Distribution of Year

# checking out the Year column
sales_data %>% 
  group_by(Year) %>% 
  count() %>% 
  ggplot() + 
  geom_density(aes(Year)) + scale_x_continuous(breaks=number_ticks(12)) + ggtitle('Global Sales by Year')

C. Understanding the Distribution of Platform

# Sales by platform :
sales_data %>% group_by(Platform) %>% 
  dplyr::summarise(sales = sum(Global_Sales)) %>% ggplot() + 
  geom_bar(aes(reorder(Platform, sales), sales), stat = "identity", 
           fill = "steelblue") + 
  xlab("Platform") + ylab("Global sales") + scale_y_continuous(breaks=number_ticks(5)) + ggtitle('Global Sales by Platform') +
  coord_flip() 

# combining the platforms together to get a better representation
sales_data = sales_data %>% 
  mutate(platform2 = case_when(
  Platform %in% c("Wii", "DS", "3DS", "WiiU", "GC", "GB","GBA") ~ "Nintendo",
  Platform %in% c("X360", "XB", "XOne") ~ "XBox",
  Platform %in% c("PS3", "PS4", "PS2", "PS", "PSP", "PSV") ~ "PS",
  Platform == "PC" ~ "PC",
  Platform == "DC" ~ "Sega"
))

sales_data %>% group_by(platform2) %>% 
  dplyr::summarise(sales = sum(Global_Sales)) %>% ggplot() + 
  geom_bar(aes(reorder(platform2, sales), sales), stat = "identity", 
           fill = "steelblue") + 
  xlab("Platform") + ylab("Global sales") + scale_y_continuous(breaks=number_ticks(5)) + ggtitle('Global Sales for Most Popular Platforms') +
  coord_flip()

Inference:
From the above barplots we can observe the top 3 most popular platforms(most sales over the years) for playing games are PS2, XBox 360 and PS3 respectively. But when we group by the manufacturer of the consoles and count the aggregate sales, we find that Play Station is the top performer while XBox has lowe sales than Nintendo’s consoles, and Sega has the lowest of all.

D. Exploring Genre : Which Genre has most of the sales?

ggplot(sales_data, aes(x=Genre, y=Global_Sales, fill=Genre, color=Genre)) +
  geom_bar(stat="identity") + 
  theme(axis.text.x=element_text(size=10, angle=90), legend.position = 'bottom', aspect.ratio=10/15, legend.key.size = unit(0.3, 'cm')) +
  ggtitle('Overall Sales by Genre')

D. Exploring Sales by Publisher : Which publisher has the highest Global Sales

# Count occurrences of each value in the Col1 column
value_counts <- table(sales_data$Publisher)

# Extract values with a count of 1
values_with_count_1 <- names(value_counts[value_counts > 411])


# Filter the dataframe based on values with count 1 in Col1 column
filtered_df <- sales_data %>%
  filter(Publisher %in% values_with_count_1)

ggplot(filtered_df, aes(x=Publisher, y=Global_Sales, fill=Publisher)) + theme(legend.position="bottom", axis.text.y=element_text(size=8, angle=45), legend.key.size = unit(0.3, 'cm')) +
  geom_bar(stat="identity") + ggtitle("Overall Sales by Publisher") + coord_flip()

Inference:
From the above two barplots, we can observe that the Genres Action, Shooter and Sports contribute higer to the total sales while lowest purchases games belongs to genres like Puzzle, Strategy and Adventure.
And Publishers like Nintendo, Electronic Arts and Activision seems to be the market leaders in terms of Total sales over all of the years.

E. Pair Plot

# Select only numeric columns
num_cols = sapply(sales_data, is.numeric)
df_num = sales_data[, num_cols]

# Create ggpairs for numeric columns
ggpairs(df_num, title='Pair Plot')

Inference:
As anticipated, [NA, EU, JP, Other]Sales are highly correlated with Global Sales and each other which suggests that if a newly launched game has comparatively higher sales than its competitors in one region, then launching the same game in other regions might result in comparatively similar trend probablistically.
Year is not significantly correlated with any other attribute

F. Understanding the relation of No of games released with Global Sales

color <- c("Titles released" = "red", "Global sales" = "blue")

sales_data %>% 
  group_by(Year) %>% 
  dplyr::summarise(sales = sum(Global_Sales), count = n())%>%
  
  ggplot() + geom_line(aes(Year, count, group = 1, color = "Titles released")) + 
  
  geom_line(aes(Year, sales, group = 1, color = "Global sales")) + 
  
  ylab("Titles released") +
  
  scale_color_manual(values = color) + labs(color = "") + scale_x_continuous(breaks=number_ticks(12)) + ggtitle('Titles Released and Sales by Year')

Inference:

SALES DATA

# Log transforming the sales column to check the relationship
ggplot(sales_data, aes(x = Year, y = log10(Global_Sales))) + geom_point()

colnames(sales_data)

##  [1] "Rank"         "Name"         "Platform"     "Year"         "Genre"       
##  [6] "Publisher"    "NA_Sales"     "EU_Sales"     "JP_Sales"     "Other_Sales" 
## [11] "Global_Sales" "platform2"

table(sales_data$Year)

## 
## 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 
##    9   46   36   17   14   14   21   16   15   17   16   41   43   60  121  219 
## 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 
##  263  289  379  338  349  482  829  775  763  941 1008 1202 1428 1431 1259 1139 
## 2012 2013 2014 2015 2016 2017 2020 
##  657  546  582  614  344    3    1

#sales_data[sales_data$Year > 2016, ]

# Since there is not enough data past 2016, we shall limit our analysis only till 2016
sales_data = sales_data[sales_data$Year < 2017, ]

sales_by_year = aggregate(cbind(NA_Sales, EU_Sales, JP_Sales, Other_Sales, Global_Sales) ~ Year, data = sales_data, sum)

2. Multivariate Analysis :

Line-Plots of the sales columns :

# Area comparison
ggplot(sales_by_year, aes(x = Year)) +
  geom_line(aes(y = NA_Sales, color = "NA_Sales")) +
  geom_line(aes(y = EU_Sales, color = "EU_Sales")) +
  geom_line(aes(y = JP_Sales, color = "JP_Sales")) +
  geom_line(aes(y = Other_Sales, color = "Other_Sales")) +
  geom_line(aes(y = Global_Sales, color = "Global_Sales")) +
  labs(x = "Year", y = "Sales", color = "Sales Type") +
  ggtitle("Sales by Year") +
  scale_color_manual(values = c("NA_Sales" = "red", "EU_Sales" = "blue", 
                                 "JP_Sales" = "green", "Other_Sales" = "orange", 
                                 "Global_Sales" = "black")) + theme(legend.position = 'bottom', aspect.ratio=10/15) + scale_x_continuous(breaks=number_ticks(12))

# Reshape data to long format
sales_by_year_long <- reshape2::melt(sales_by_year, id.vars = "Year", variable.name = "Sales_Location", value.name = "Sales")
#head(sales_by_year_long)
sales_by_year_short <- sales_by_year_long[sales_by_year_long$Sales_Location != "Global_Sales", ]
ggplot(data = sales_by_year_short) +
  geom_line(aes(x = Year, y = Sales, colour = Sales_Location, group=Sales_Location)) + scale_x_continuous(breaks=number_ticks(12)) +
  ggtitle('Reginal Sales Over the Years')

sales_data %>% group_by(platform2, Year) %>%
  dplyr::summarise(sales = sum(Global_Sales)) %>% 
  ggplot() + 
  geom_line(aes(Year, sales, group = platform2, color = platform2)) +
  xlab("Year of release") + ylab("Global Sales") + labs(color = "") + ggtitle('Global Sales for each Platform')

## `summarise()` has grouped output by 'platform2'. You can override using the
## `.groups` argument.

# Top 3 best sellers overall
global_sales_sorted <- sales_data[order(-sales_data$Global_Sales), ]
top_global_sales <- global_sales_sorted[1:3, c("Name", "Platform", "Global_Sales")]

print(top_global_sales)

##                Name Platform Global_Sales
## 1        Wii Sports      Wii        82.74
## 2 Super Mario Bros.      NES        40.24
## 3    Mario Kart Wii      Wii        35.82

Understanding Game, Platform Popularity in North America

# NA Subset and sort
na_sales <- sales_data[!is.na(sales_data$NA_Sales), ]
na_sales_sorted <- na_sales[order(-na_sales$NA_Sales), ]

# top 10 with highest sales
top_na_sales <- na_sales_sorted[1:10, c("Name", "Platform", "NA_Sales")]


t_v_name_NA <- aggregate(list(NA_Sales = sales_data$NA_Sales), list(Name = sales_data$Name), sum)
t_v_name_NA <- t_v_name_NA[order(t_v_name_NA$NA_Sales, decreasing = T), ]

ggplot(data = top_na_sales, aes(x = Name, y = NA_Sales, fill = Name)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "NA Sales (in millions)", 
       title = "Top 10 Best Sellers in North America") + 
  theme(axis.text.y=element_text(size=7, angle=45), legend.position = 'none') + coord_flip()

# NA Subset and sort
na_sales <- sales_data[!is.na(sales_data$NA_Sales), ]
na_sales_sorted <- na_sales[order(-na_sales$NA_Sales), ]

# top 10 with highest sales
top_na_sales <- na_sales_sorted[1:10, c("Name", "Platform", "NA_Sales")]


t_v_name_NA <- aggregate(list(NA_Sales = sales_data$NA_Sales), list(Name = sales_data$Name), sum)
t_v_name_NA <- t_v_name_NA[order(t_v_name_NA$NA_Sales, decreasing = T), ]

ggplot(data = top_na_sales, aes(x = Name, y = NA_Sales, fill = Platform)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "NA Sales (in millions)", 
       title = "Top 10 Best Sellers in North America") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Game, Platform Popularity in Europe

# EU Subset and sort
eu_sales <- sales_data[!is.na(sales_data$EU_Sales), ]
eu_sales_sorted <- eu_sales[order(-eu_sales$EU_Sales), ]

# top 10 with highest sales
top_eu_sales <- eu_sales_sorted[1:10, c("Name", "Platform", "EU_Sales")]

ggplot(data = top_eu_sales, aes(x = Name, y = EU_Sales, fill = Platform)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "EU Sales (in millions)", 
       title = "Top 10 Best Sellers in Europe") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Game, Platform Popularity in Japan

# JP Subset and sort
jp_sales <- sales_data[!is.na(sales_data$JP_Sales), ]
jp_sales_sorted <- jp_sales[order(-jp_sales$JP_Sales), ]

# top 10 with highest sales
top_jp_sales <- jp_sales_sorted[1:10, c("Name", "Platform", "JP_Sales")]

ggplot(data = top_jp_sales, aes(x = Name, y = JP_Sales, fill = Platform)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "JP Sales (in millions)", 
       title = "Top 10 Best Sellers in Japan") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Game, Platform Popularity in Other Regions

# JP Subset and sort
o_sales <- sales_data[!is.na(sales_data$Other_Sales), ]
o_sales_sorted <- o_sales[order(-o_sales$Other_Sales), ]

# top 10 with highest sales
top_o_sales <- o_sales_sorted[1:10, c("Name", "Platform", "Other_Sales")]

ggplot(data = top_o_sales, aes(x = Name, y = Other_Sales, fill = Platform)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "Other Sales (in millions)", 
       title = "Top 10 Best Sellers in Other Regions") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Genre Popularity in North America

# NA Subset and sort
na_sales <- sales_data[!is.na(sales_data$NA_Sales), ]
na_sales_sorted <- na_sales[order(-na_sales$NA_Sales), ]

# top 10 with highest sales
top_na_sales <- na_sales_sorted[1:10, c("Name", "Platform", "NA_Sales", "Genre")]


t_v_name_NA <- aggregate(list(NA_Sales = sales_data$NA_Sales), list(Name = sales_data$Name), sum)
t_v_name_NA <- t_v_name_NA[order(t_v_name_NA$NA_Sales, decreasing = T), ]

ggplot(data = top_na_sales, aes(x = Name, y = NA_Sales, fill = Genre)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "NA Sales (in millions)", 
       title = "Top 10 Best Sellers in North America ~ Genre") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Genre Popularity in Europe

# EU Subset and sort
eu_sales <- sales_data[!is.na(sales_data$EU_Sales), ]
eu_sales_sorted <- eu_sales[order(-eu_sales$EU_Sales), ]

# top 10 with highest sales
top_eu_sales <- eu_sales_sorted[1:10, c("Name", "Platform", "EU_Sales", "Genre")]

ggplot(data = top_eu_sales, aes(x = Name, y = EU_Sales, fill = Genre)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "EU Sales (in millions)", 
       title = "Top 10 Best Sellers in Europe ~ Genre") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Genre Popularity in Japan

# JP Subset and sort
jp_sales <- sales_data[!is.na(sales_data$JP_Sales), ]
jp_sales_sorted <- jp_sales[order(-jp_sales$JP_Sales), ]

# top 10 with highest sales
top_jp_sales <- jp_sales_sorted[1:10, c("Name", "Platform", "JP_Sales", "Genre")]

ggplot(data = top_jp_sales, aes(x = Name, y = JP_Sales, fill = Genre)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "JP Sales (in millions)", 
       title = "Top 10 Best Sellers in Japan ~ Genre") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Understanding Genre Popularity in Other Regions

# JP Subset and sort
o_sales <- sales_data[!is.na(sales_data$Other_Sales), ]
o_sales_sorted <- o_sales[order(-o_sales$Other_Sales), ]

# top 10 with highest sales
top_o_sales <- o_sales_sorted[1:10, c("Name", "Platform", "Other_Sales", "Genre")]

ggplot(data = top_o_sales, aes(x = Name, y = Other_Sales, fill = Genre)) + 
  geom_bar(stat = "identity", color='white') +
  labs(x = "Game Title", y = "Other Sales (in millions)", 
       title = "Top 10 Best Sellers in Other Regions ~ Genre") + 
  theme(axis.text.y=element_text(size=8, angle=45), legend.position = 'bottom') + coord_flip()

Top Publisher for each year

years = unique(sales_data$Year)
df_top_p = data.frame(Year = integer(), 
                      Publisher = character(), 
                      Global_Sales = numeric(), 
                      stringsAsFactors = FALSE)

for (year in years) {
  # games released for current year
  year_sales <- subset(sales_data, Year == year)
  # Aggregate sales
  publisher_sales <- aggregate(Global_Sales ~ Publisher, year_sales, sum)
  
  # sory by descending and take the top
  top_publisher <- publisher_sales[order(-publisher_sales$Global_Sales), "Publisher"][1]
  # add into newdf
  df_top_p <- rbind(df_top_p, data.frame(Year = year, Publisher = top_publisher, Global_Sales = publisher_sales[publisher_sales$Publisher == top_publisher, "Global_Sales"]))
}

ggplot(data = df_top_p, aes(x = Year, y = Global_Sales, fill = Publisher)) +
  geom_col(position = "dodge") +
  labs(x = "Year", y = "Global Sales (in millions)", 
       title = "Top Publisher by Global Sales for Each Year") +
  theme(axis.text.x=element_text(size=8, angle=90), legend.position = 'bottom', legend.key.size = unit(0.2, 'cm')) + scale_x_continuous(breaks=number_ticks(21))

Inference:
As we can see from the above plot, Nintendo was the top Publisher for most of the years. How ever we can see a trend where Electronic Arts started selling more games than Nintendo from year 2002 until 2005 after which again Nintendo was top publisher. And lastly from the year 2010 Electronic Art’s sales were more than Nintendo.
Hence We are particularly interested in analyzing data from years 2005-2009 and 2010-2014 which will help us understand how Electronic Arts was able to beat it competitor Nintendo.

Further Analysis from year 2005-2009 [mid 5 years] and year 2010-2014[last 5 years]

#head(filtered_df)
mid5y = filtered_df %>% filter(Year>=2005, Year<=2009)
last5y = filtered_df %>% filter(Year>=2010, Year<=2014)
last5y

Checking out the Sales for mid 5 years

ggplot(mid5y, aes(x=Year, y=Global_Sales))+
  geom_bar(stat='identity', fill="navy")+
  facet_wrap(~Publisher, nrow=2) + ggtitle('Sales from year 2005-2009 ~ Publisher') + theme(axis.text.x=element_text(size=9, angle=90))

ggplot(filter(mid5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo'), aes(x=Year, y=Global_Sales))+
  geom_bar(stat='identity', fill="navy")+
  facet_wrap(~Publisher, nrow=1) + ggtitle('Sales from 2005-2009 ~ Publisher') + theme(axis.text.x=element_text(size=8, angle=90))

ggplot(last5y, aes(x=Year, y=Global_Sales))+
  geom_bar(stat='identity', fill="navy")+
  facet_wrap(~Publisher, nrow=2) + ggtitle('Sales from year 2010-2014 ~ Publisher') + theme(axis.text.x=element_text(size=9, angle=90))

ggplot(filter(last5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo'), aes(x=Year, y=Global_Sales))+
  geom_bar(stat='identity', fill="navy")+
  facet_wrap(~Publisher, nrow=1) + ggtitle('Sales from 2010-2014 ~ Publisher') + theme(axis.text.x=element_text(size=8, angle=90))

Inference:
From the above two faceted plots, we can see that Nintendo, Electronic Arts and Acticision are having the best sales
In year 2005-2009 Nintendo sales were all time high about 200 million and averaged around 130 million over the period of 5 years, while Electronic Arts had average sales of 75 million followed by Activision with 60 million in sales
However, from year 2010 Nintendo’s sales dropped to average 55 million and Electronic Art’s average sales were highest around 60 million followed by Activision with average of 50 million.
In summary, we saw that Nintendo was having highest sales in mid-5-years while in the last-5-years Electronic Arts crossed Nintendo in average sales and Activision was also able to increase the sales comparatively.
We are interested in finding the probable reason why Nintendo’s sales dropped by around 50% while sales for Electronic Arts and Activision increased.

For number of game relased by publishers

count_last5y <- last5y %>%
  group_by(Publisher, Genre, Year) %>%
  dplyr::summarise(count = n())

## `summarise()` has grouped output by 'Publisher', 'Genre'. You can override
## using the `.groups` argument.

count_last5y

count_mid5y <- mid5y %>%
  group_by(Publisher, Genre, Year) %>%
  dplyr::summarise(count = n())

## `summarise()` has grouped output by 'Publisher', 'Genre'. You can override
## using the `.groups` argument.

#filter(count_last5y, Publisher=='Activision')[,2:4]

ggplot(data = count_mid5y, aes(x=Genre, y=count, fill=Publisher)) + 
    geom_bar(stat="identity") +
  facet_wrap('Year', nrow=1) + ylim(0,150) + 
  theme(axis.text.x=element_text(size=9, angle=90), legend.position = 'bottom', legend.key.size = unit(0.2, 'cm')) + ggtitle('Games Released for each Genre ~ mid-5-Years')

ggplot(data = count_last5y, aes(x=Genre, y=count, fill=Publisher)) + 
    geom_bar(stat="identity") +
  facet_wrap('Year', nrow=1) + ylim(0,150) + 
  theme(axis.text.x=element_text(size=9, angle=90), legend.position = 'bottom', legend.key.size = unit(0.3, 'cm')) + ggtitle('Games Released for each Genre ~ last-5-Years')

Inference:
The above plots are hard to interpret but we can spot a trend that the relatively higher number of games are realeased for Action category
To better understand similar kind of plots we will only consider publishers like “Nintendo”, “Activision” and “Electronic Arts” since we spotted a pattern for these publishers in previous plots for this set of years

Grouped Bar Plot for the top 3 publishers and all Genres

count_mid5y <- mid5y %>%
  group_by(Publisher, Genre, Year) %>%
  dplyr::summarise(count = n())

## `summarise()` has grouped output by 'Publisher', 'Genre'. You can override
## using the `.groups` argument.

count_mid5y

Grouped Bar Plot for mid-5-years

ggplot(data = filter(count_mid5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo'), aes(x=Year, y=count, fill=Publisher, group=Publisher)) + 
    geom_bar(stat="identity", position="dodge") +
  facet_wrap('Genre', nrow=1) + 
theme(axis.text.x=element_text(size=8, angle=90), legend.key.size = unit(0.2, 'cm')) + ggtitle('Games released by Publishers(2005-2009) ~ Genre') + ylim(0,40)

## Warning: Removed 5 rows containing missing values (`geom_bar()`).

Focusing into Genres with noticable trends mid-5-years

m1 = filter(count_mid5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo')
m2 = filter(m1, Genre=='Action' | Genre=='Adventure' | Genre=='Puzzle' | Genre=='Shooter' | Genre=='Sports' | Genre=='Racing')

ggplot(data = filter(m2, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo'), aes(x=Year, y=count, fill=Publisher, group=Publisher)) + 
    geom_bar(stat="identity", position="dodge") +
  facet_wrap('Genre', nrow=1) + 
  theme(axis.text.x=element_text(size=8, angle=90), legend.key.size = unit(0.2, 'cm')) + ggtitle('Games released by Publishers(2005-2009) ~ Genre') + ylim(0,40)

## Warning: Removed 5 rows containing missing values (`geom_bar()`).

Focusing into Genres with noticable trends last-5-years

l1 = filter(count_last5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo')
l2 = filter(l1, Genre=='Action' | Genre=='Adventure' | Genre=='Puzzle' | Genre=='Shooter' | Genre=='Sports' | Genre=='Racing'| Genre=='Misc')

ggplot(data = l2, aes(x=Year, y=count, fill=Publisher, group=Publisher)) + 
    geom_bar(stat="identity", position="dodge") +
  facet_wrap('Genre', nrow=1) + 
  theme(axis.text.x=element_text(size=8, angle=90), legend.key.size = unit(0.2, 'cm')) + ggtitle('Games released by Publishers(2010-2014) ~ Genre') + ylim(0,40)

Inference:
We can see that grouped barplot is not a good representation hence using stacked barplot to better understand the trends for game relases over the same set of years

Stacked Bar Plots for top 3 Publishers for mid-5-years with Genres in which most of the games were released

m1 = filter(count_mid5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo')
m2 = filter(m1, Genre=='Action' | Genre=='Puzzle' | Genre=='Shooter' | Genre=='Sports' | Genre=='Racing' | Genre=='Misc')

ggplot(data = m2, aes(x=Year, y=count, fill=Publisher)) + 
    geom_bar(stat="identity") +
  facet_wrap('Genre', nrow=2) + 
  theme(axis.text.x=element_text(size=10, angle=90), legend.key.size = unit(0.2, 'cm'), legend.position = 'bottom') + ggtitle('Games released by Publishers(2005-2009) ~ Genre') + ylim(0,40)

## Warning: Removed 5 rows containing missing values (`position_stack()`).

## Warning: Removed 3 rows containing missing values (`geom_bar()`).

l1 = filter(count_last5y, Publisher=='Activision' | Publisher=='Electronic Arts' | Publisher=='Nintendo')
l2 = filter(l1, Genre=='Action' | Genre=='Puzzle' | Genre=='Shooter' | Genre=='Sports' | Genre=='Racing'| Genre=='Misc')

ggplot(data = l2, aes(x=Year, y=count, fill=Publisher)) + 
    geom_bar(stat="identity") +
  facet_wrap('Genre', nrow=2) + 
  theme(axis.text.x=element_text(size=10, angle=90), legend.key.size = unit(0.2, 'cm'), legend.position = 'bottom') + ggtitle('Games released by Publishers(2010-2014) ~ Genre') + ylim(0,40)

## Warning: Removed 3 rows containing missing values (`geom_bar()`).

Modelling:

sales_by_publisher = sales_data %>% group_by(Publisher) %>% 
  dplyr::summarise(sales = sum(Global_Sales)) %>% 
  arrange(desc(sales)) %>% slice(1:10)

top_publishers = head(sales_by_publisher$Publisher, n = 10)

filtered_df = sales_data %>%filter(Publisher %in% top_publishers)

sales_by_year_publisher = aggregate(Global_Sales ~ Publisher + Year, data = filtered_df, sum)

ggplot(sales_by_year_publisher, aes(x = Year, y = Global_Sales)) +
    geom_point(alpha=0.75,size=0.5) +
    geom_smooth(method = "loess", se = FALSE, span=0.75, color= 'blue', linewidth= 0.6) + 
  facet_wrap(~ Publisher, ncol = 5) +
  labs(title = "Trend of Total Global Sales by Publisher and Year",
       x = "Year",
       y = "Total Global Sales") +
      theme(axis.text.x=element_text(size=9))

## `geom_smooth()` using formula = 'y ~ x'

#geom_smooth(method = "lm",color= 'red', se =FALSE,linewidth= 0.8) +

sales_by_genre_year_publisher = aggregate(Global_Sales ~ Genre + Year, data = filtered_df, sum)
ggplot(sales_by_genre_year_publisher, aes(x = Year, y = Global_Sales)) +
  geom_point(alpha=0.75,size=0.5) +
    geom_smooth(method = "loess", se = FALSE, span=0.75, color= 'blue', linewidth= 0.6) +
  facet_wrap(~ Genre, ncol = 4) +
  labs(title = "Total Global Sales by Genre and Year",
       x = "Year",
       y = "Total Global Sales") +
  theme(axis.text.x=element_text(size=9))

## `geom_smooth()` using formula = 'y ~ x'

Modelling:

sales_by_publisher = sales_data %>% group_by(Publisher) %>% 
  dplyr::summarise(sales = sum(Global_Sales)) %>% 
  arrange(desc(sales)) %>% slice(1:10)

top_publishers = head(sales_by_publisher$Publisher, n = 10)

filtered_df = sales_data %>%filter(Publisher %in% top_publishers)

sales_by_year_publisher = aggregate(Global_Sales ~ Publisher + Year, data = filtered_df, sum)

ggplot(sales_by_year_publisher, aes(x = Year, y = Global_Sales)) +
    geom_point(alpha=0.75,size=0.5) +
    geom_smooth(method = "loess", se = FALSE, span=0.75, color= 'blue', linewidth= 0.6) + 
  geom_smooth(method = "lm",color= 'red', se =FALSE,linewidth= 0.8) +
  facet_wrap(~ Publisher, ncol = 5) +
  labs(title = "Trend of Total Global Sales by Publisher and Year",
       x = "Year",
       y = "Total Global Sales") +
      theme(axis.text.x=element_text(size=9))

## `geom_smooth()` using formula = 'y ~ x'
## `geom_smooth()` using formula = 'y ~ x'

sales_by_genre_year_publisher = aggregate(Global_Sales ~ Genre + Year, data = filtered_df, sum)
ggplot(sales_by_genre_year_publisher, aes(x = Year, y = Global_Sales)) +
  geom_point(alpha=0.75,size=0.5) +
    geom_smooth(method = "loess", se = FALSE, span=0.75, color= 'blue', linewidth= 0.6) + 
 geom_smooth(method = "lm",color= 'red', se =FALSE,linewidth= 0.8) +
  facet_wrap(~ Genre, ncol = 4) +
  labs(title = "Total Global Sales by Genre and Year",
       x = "Year",
       y = "Total Global Sales") +
  theme(axis.text.x=element_text(size=9))

## `geom_smooth()` using formula = 'y ~ x'
## `geom_smooth()` using formula = 'y ~ x'